[gmx-developers] gmx cluster improuvement

[David van der Spoel]

Den 2018-07-26 kl. 13:05, skrev Boris Timofeev:
> Dear developers!
> 
> In processing of results of REMD there were inconveniences which, I think,
> it will be easy to eliminate in future versions of Gromacs.
> 
This is a useful suggestions, however it is not entirely trivial to 
choose which box one would use. Please start a redmine!


> 1. Command
> gmx cluster ... -cl somefile.pdb ...
> unlike command
> gmx trjconv ... -o somefile.pdb ...
> doesn't form record about the box sizes in the PDB file:
> CRYST1 boxx boxy boxz ...
> 
> As a result, subsequent call
> gmx hbond .. -f somefile.pdb ....
> deals to the error:
> 
> Program: gmx hbond, version 2016.2
> Source file: src\gromacs\gmxana\gmx_hbond.cpp (line 947)
> Fatal error:
> Your computational box has shrunk too much.
> gmx hbond can not handle this situation, sorry.
> 
> Manual addition the record CRYST1 to pdb-file solves this problem.
> I suggest to add formation of the record CRYST1 to output pdb,
> formed by the gmx cluster command.
> 
> 2. The gmx trjconv command can select parts of a trajectory by means of 
> the index file: - sub index.ndx:
> 
> "Use option -sub. This assumes that the entries in the index file are 
> frame numbers and
> dumps each group in the index file to a separate trajectory file."
> 
> It would be very convenient if the "gmx cluster" command could form,
> optionally, index files for the found clusters.
> 
> Now this problem is solved by copying of times from gmx cluster 
> command's log-file
> and transformation of times to indexes of frames manually. This is very 
> uncomfortable.
> 
> Sincerely yours,
> Boris Timofeev
> 
> 
> 
> 
> -- 
> Boris Timofeev
> 
> 


-- 
David van der Spoel, Ph.D., Professor of Biology
Head of Department, Cell & Molecular Biology, Uppsala University.
Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
http://www.icm.uu.se